Frederik Treue <frtr at fysik.dtu.dk> writes: > I'm trying to solve a convection-diffusion problem. The method I use > requires me to solve some Helmholtz and Poisson equations, which is the > time consuming part. In order to reduce this time, I try to use > preconditioners. However, no matter which preconditioner I use (I've > tried PCJACOBI,PCBJACOBI,PCPBJACOBI,PCMG with 2 levels, galerkin) it > introduces errors along the edges of the local domains when using > multiple processors. These errors are small, but they don't converge to > 0 as a function of resolution.
How "small"? How are you measuring the error? Does it converge to match the internal error as you shrink -ksp_rtol? It should not happen with -pc_type jacobi (or pbjacobi) because they don't do anything special at boundaries.
